About

I’m Ryan Lowe. I architect cloud systems, and for the past year I’ve been building agents that run on top of them.

I live in Rotorua, New Zealand. Twenty years in, I still write code. The case studies on this site are working systems, not slideware.

1.0Where the architecture comes from

Three years at AWS as a solutions architect was the formative stretch. I worked on Amazon Pinpoint and SES, designing serverless, event-driven, multi-tenant reference systems and publishing them (AWS Solutions, reference architectures, blog posts, videos; indexed in the archive), then sat with a few hundred engineering teams while they argued about their own designs. You learn a particular thing doing that. Not how one system works, but which patterns keep failing, and why. Later I joined the Alexa Skills Kit team and watched developers build voice-first conversational products, which turned out to be a fairly direct rehearsal for what everyone is building now.

Before AWS I founded and ran Centric Consulting’s Seattle office, growing it to US$20M and a hundred people in three years, much of it embedded at Microsoft designing marketing platforms built to scale to hundreds of millions of users. Before that, ExactTarget when it was still ExactTarget, and a decade of engineering jobs where I was the one on call.

2.0Microsoft

I’ve been at Microsoft since 2022: first as an employee, and since moving to New Zealand in 2025 as an independent contractor. I lead technical program and product work across a portfolio of data capabilities behind Microsoft’s marketing engine: the roadmap, the technical requirements, the architecture decisions, and the work of turning what a marketing organisation says it needs into something engineers can build. Four technical program managers report to me; the building is done by engineering teams across three vendor agencies. I’m still in the Python and the SQL most weeks, which is the part I’d miss.

The AI work I’ve done there is smaller than my title might imply, and I’d rather say so plainly. I got the organisation’s first generative-AI pilot funded and delivered, took prototypes through Responsible AI review, and set my team’s practice for AI-assisted delivery. Useful work. Not the same thing as running an AI platform.

3.0SplitKit

The agentic engineering is SplitKit: a production platform for race-timing companies that I designed, built and operate on my own. No team.

It isn’t a venture I’m pitching. It’s where the practice stays current: external API ingestion, per-participant records, event-triggered SMS and email, generated finisher narratives, validation gates before anything reaches a runner’s phone, cost controls, and the unglamorous business of operating something you shipped yourself. Most of the writing on this site comes out of it.

It grew out of RaceResults360, which I co-founded in 2012: mobile apps, live results and on-site systems for the NYC Triathlon, Escape from Alcatraz, Life Time Triathlon and a few dozen others. Better timing technology eventually displaced us and we closed it cleanly. I learned more from that than from most things that worked.

4.0What I’m interested in now

Most AI projects I see stall in the same place. The model is fine. What’s missing is everything around it: grounding in data someone actually trusts, evaluation, something checking the output before a customer sees it, an architecture that holds up on a bad day. That gap is where I’m useful, and it’s what most of the writing here is about.

The case studiesare the best place to start. If you’d like to talk: hi@ryanjlowe.com